Structural Bioinformatics Subproject We developed a series of structural bioinformatics tools, including the SkyLine homology modeling pipeline, the SkyBase database of homology models and the MarkUS function annotation server. Our plan in the coming years is to continue in the development of the various software tools and to apply them broadly. In terms of technology, our specific plans include: a) The continued development of Skybase and its extension to cover models derived from all proteins of known structure, b) The continued develop of MarkUs including the addition of new analysis tools that predict protein-ligand and protein-DNA interactions, c) The creation of a database which combines SkyBase structures with MarkUs-derived function annotation. Entries in this databse will be linked to a variety of genomic databases, e.g., expression or high-throughput protein-protein interaction databases. The combination of SkyBase and MarkUs renders a powerful framework for discerning protein families and for characterizing the multi-dimensional relationships among them, with obvious implications for genome-scale biology, d) The continued integration of our tools with the PSI Knowledge Base. We will provide ongoing function annotation and leverage analysis for all PSI structures so that our databases will be continually expanded. In addition, we will create models and functional inferences for proteins involved in cancer-related networks. We have demonstrated in our work on START domains that by focusing on specific protein families, building models for all family members and organizing structural and functional date into a family-specific database that we are able to derive new functional inferences and identify new family members. The general strategy involved in family analysis will be implemented for a set of proteins chosen from cancer-related networks of interest to NESG. In addition, we will begin in the construction of models of complexes involved in these networks. A database of models for complexes will be created based on an expansion of the methods used to create Skybase and on methods being developed in the Honig lab to predict protein-protein binding interfaces. The models will be evaluated based on a variety of sequence, structure and biophysical properties and on an analysis of the biological literature. This type of effort is labor intensive but can yield enormous payoffs both in terms of biological understanding and in guiding experimental efforts.